INDEX

Explanations

specific terms related to events, organizations, and names

oai_token-act-pair · gpt-3.5-turbo

references to events, media, and organizational titles

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

âĶĢâĶĢ

-0.59

ioch

-0.57

kefeller

-0.55

 sidx

-0.55

 stalls

-0.54

ÃĥÃĤ

-0.53

arching

-0.53

etheus

-0.52

 toget

-0.52

enson

-0.51

POSITIVE LOGITS

 Destruction

0.64

ibrary

0.63

 Emails

0.61

 Meditation

0.61

 Deaths

0.61

 Suff

0.59

 Mood

0.59

 Seraph

0.58

 Diseases

0.57

 abound

0.56

Activations Density 1.032%

specific terms related to events, organizations, and names

references to events, media, and organizational titles

No Comments

No Known Activations